Upload
leonardo-ruivo
View
221
Download
0
Embed Size (px)
Citation preview
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 1/22
Can Science Know When You're Conscious?
Page 1 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
University Press Scholarship Online
Oxford Scholarship Online
Pathways to Knowledge: Private and PublicAlvin I. Goldman
Print publication date: 2002
Print ISBN-13: 9780195138795
Published to Oxford Scholarship Online: November 2003
DOI: 10.1093/0195138791.001.0001
Can Science Know When You're Conscious?
Epistemological Foundations of Consciousness Research
Alvin I. Goldman (Contributor Webpage)
DOI:10.1093/0195138791.003.0006
Abstract and Keywords
Consciousness researchers commonly rely on their subjects’ verbal reports to
determine their conscious states. Is this defensible in the conduct of science? Attempts
might be made to rationalize the reliance on verbal reports by appealing to higher‐order
thought or functionalist approaches to consciousness, but these are rejected. A third
approach is defended, based on subjects’ introspective capacities. Admittedly, the
reliability of introspection cannot be independently validated, but an independent
validation requirement is too restrictive for an epistemologically “basic” method.
Keywords: consc iousness, epistemology, functionalism, higher‐order thought, introspection, verbal report
Consciousness researchers standardly rely on their subjects' verbal reports to
ascertain which conscious states they are in. What justifies this reliance on verbal
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 2/22
Can Science Know When You're Conscious?
Page 2 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
reports? Does it comport with the third‐person approach characteristic of science,
or does it ultimately appeal to first‐person knowledge of consciousness? If first‐
person knowledge is required, does this pass scientific muster? Several attempts
to rationalize the reliance on verbal reports are considered, beginning with
attempts to define consciousness via the higher‐order thought approach and
functionalism. These approaches are either (A) problematic in their own right, or (B)
ultimately based on a first‐person access to consciousness. A third approach
assumes that scientists can trust verbal reports because subjects reliably monitor
or ‘introspect’ their conscious states. This raises the question of whether the
reliability of introspection (or self‐monitoring) can be validated by independent
criteria. Merikle's attempts to validate this reliability are shown to involve some
unavoidable circularity. It is conjectured that scientists' reliance on their subjects'
verbal reports tacitly appeals to their own introspective reliability, which is not
independently validatable. Some epistemologists might conclude that this renders
scientists' conclusions about conscious states unjustified, but I argue that this does
not contravene the constraints of a proper epistemology.
1. Why Rely on Verbal Reports?Of the many psychological states people occupy, some are conscious and some are not;
that is, some involve awareness and some do not. (In what follows, I use the terms
“conscious” and “aware” interchangeably.) 1 Scientific (p.115) research on
consciousness seeks to determine which types of states tend to be conscious and why.
To answer these questions, consciousness researchers need to ascertain the presence
or absence of conscious states within their subjects on specific occasions. How do
researchers do that given that their subjects' conscious states are not directly
observable? The answer, of course, is that scientists observe their subjects' behaviour,
especially their verbal behaviour, and use that behaviour to infer the presence or
absence of conscious states. This looks like familiar scientific procedure. By observing
public behaviour, consciousness can be studied in an objective, third‐person fashion,
much like any other topic in cognitive science or neuroscience. This is the standard
methodological view, I suspect, among scientific researchers of consciousness. Here I
want to ask whether this view is correct. Can consciousness be studied scientifically in a
purely third‐person fashion, without relying, perhaps tacitly, on first‐person knowledge or
warrant? If first‐person knowledge does turn out to be required, is the methodological
respectability of consciousness research thereby threatened? If its epistemological viability is threatened, can this threat be met or overcome?
My discussion focuses on verbal reports. Verbal reports obviously play a central role in
most methodologies of consciousness. For example, how do consciousness researchers
determine that patients with blindsight lack visual awareness in certain portions of their
visual field? Patients report the absence of such awareness, and researchers accept their
reports. How do memory researchers determine that certain subjects lack “explicit”
memory, i.e. conscious memory, for certain events? Again, the subjects say that they
don't remember, researchers construe such sayings as reports to the effect that no
conscious memories of the target events are present, and researchers accept these
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 3/22
Can Science Know When You're Conscious?
Page 3 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
reports as true. Admittedly, reliance on verbal reports is not wholly uncontroversial or
unqualified, especially in the case of denials of consciousness. I shall return to the reasons
for these qualifications below. In the main, however, researchers rely on their subjects'
reports as the fundamental kind of evidence for the presence or absence of conscious
states. This is clearly stated by Marcel:
There is really only one criterion for phenomenal experience. This is a person's
report, direct or indirect, that they have a sensation of one or another kind, that
they are or were conscious in one or another way. . . . Direct reports include
statements such as ‘I have a headache, or an itch’, or ‘I feel hungry’. Indirect
reports include statements such as ‘I see a light’ or ‘I hear a car’, where the
person is not directly reporting the sensation, but means by the statement that
they consciously see or hear something. . . . [P]rovided that the person is not lying,
there is little reason to doubt the validity of a report that there is phenomenal
experience. ( 1988 , 131)
What warrants consciousness researchers in so relying on verbal reports? Cognitive
psychologists and neuropsychologists would not rely, after all, on (p.116) their
subjects' reports about all psychological states or processes. When it comes to the
nonconscious sphere of mental processing—the great bulk of what transpires in the mind‐
brain—scientists would not dream of asking subjects for their opinions. Moreover, if
subjects were to offer their views about what happens (at the micro‐level) when they
parse a sentence or retrieve an episode from memory or reach for a cup, scientists
would give no special credence to these views. So what entitles scientists to rely so
heavily on subjects' reports when they concern conscious experience?
There are, I suggest, two possible types of interpretation or reconstruction of the
epistemic warrant for scientific reliance on verbal reports, a constitutive approach and a
non‐constitutive approach.
(1) Constitutive approach. The verbal report of a conscious state, or the belief
underlying such a report, is wholly or partly constitutive of there being such a
state, or of the state's being conscious. For example, part of what it might mean
for a state to be conscious is that it tends to gives rise to such a report, or to a
belief that generates such a report, or the like. So when a scientist observes such
a report and infers its underlying belief, this would provide evidence for the
presence of the reported conscious state.
(2) Non‐constitutive approach. Neither the verbal report of a conscious state nor
the belief from which it issues is (even partly) constitutive of there being such a
state or of its being conscious. However, the report and/or the belief is a reliable
indicator of the occurrence of such a state. Hence, when a scientist observes
such a report and infers its underlying belief, this provides evidence for the
presence of the reported conscious state.
To illustrate the difference between the two rationales, consider the meaning of the
phrase “run a red light.” This phrase means “drive through an intersection while the light
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 4/22
Can Science Know When You're Conscious?
Page 4 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
is red.” Given this meaning, if Jones observes the light to be red while Smith drives
through the intersection, then Jones observes something that is constitutive evidence for
Smith's running a red light. By contrast, if Jones merely hears a witness testify to the
effect that the light was red while Smith drove through the intersection, what Jones
observes is non‐constitutive evidence—at best a reliable indicator—that Smith ran a red
light. Either observation might provide adequate evidence for the proposition in question,
but the evidence is different in the two cases.
Why is there reference to an “underlying belief” in my formulation of the two
approaches? As the passage from Marcel indicates, a scientist's trust of a subject's
verbal report presumes that the subject is sincere, that she is not lying or idly mouthing
the report sentence without genuine conviction. Suppose a scientist suspects that the
subject reports a certain state of awareness (p.117) because a band of science
obstructionists offered her a reward to sabotage the study by falsifying her reports.
Such a scientist will not trust that subject's reports. In the normal case, of course,
consciousness scientists tacitly assume that a subject's reports are sincere, i.e., that she
believes what she reports. If a subject says, “I am now in conscious state M,” then she
believes that she is now in conscious state M. This presumed belief may play an important
evidential role in the scientist's conclusion, either on the constitutive or the non‐
constitutive approach. In what follows I take for granted that scientists are normally
warranted in inferring this kind of underlying belief from a subject's report without
examining the basis for such warrant. What interests me is how the scientist can justifiably
infer the actual presence or absence of a conscious state from the subject's belief in the
presence or absence of such a state.
I am not requiring, of course, that either a subject's report or her underlying belief is all
the evidence available to the scientist from which to infer a conscious state. A constitutive
approach might merely suggest that the report or the belief is partly constitutive of being
in a conscious state. Let us see whether this approach can work.
2. The HOT ApproachThe first example I consider of a constitutive approach is the higher‐order thought (HOT)
theory of consciousness, adapted for the purposes of our current problem. In its
simplest form, the HOT theory says that what it consists in for a person's mental state M
to be conscious is: (i) the person is in M, and (ii) the same person has a contemporaneous
thought, M*, about her being in M, a “higher‐order” thought. The leading defender of
the HOT theory is Rosenthal ( 1991 , 1997 , 2000 ), and I shall follow his version of it.
However, Rosenthal denies that his theory is supposed to provide a “conceptual
analysis” (or the meaning) of state‐consciousness (Rosenthal 1991 , 473). But in the
present context, where we are examining the constitutive approach, it is important that
the HOT theory be construed as a conceptual thesis, so that is the version I shall examine
here. Thus, what I shall discuss and criticize is not quite Rosenthal's view, but a very
close relative of it. 2
A further feature of the HOT approach should be emphasized. The HOT theory does not
assert that for a mental state M to be conscious, it must have a conscious meta‐thought
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 5/22
Can Science Know When You're Conscious?
Page 5 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
M*; to the contrary, even a non‐conscious meta‐thought suffices to confer consciousness
on M. The notion that a non‐conscious meta‐thought can confer consciousness on a first‐
order state is quite counter‐intuitive (in my opinion), but the HOT approach is driven to
this position by a threat of infinite regress. If a conscious meta‐thought is required to
make a lower‐order state conscious, then the existence of a conscious metathought
(p.118) implies the existence of a conscious meta‐meta‐thought, and so on ad infinitum.
This is an intolerable regress, because although people clearly have conscious states,
they do not seem to have infinite hierarchies of embedded conscious states. To avoid
such infinite hierarchies, the HOT theory is committed to allowing non‐conscious meta‐
thoughts to turn the trick of conferring consciousness on first‐order states. A further
reason why the HOT theory needs non‐conscious meta‐thoughts is that many conscious
mental states seem to lack any conscious HOTs. Consider perceptual experiences, which
are all or almost all conscious. There is conscious reflection on very few of these states,
so if what renders them conscious are HOTs, they must be non‐conscious HOTs. 3 True,
postulation of these nonconscious meta‐thoughts is inherently implausible, because it
posits an uneconomical duplication of informational effort. This is a difficulty for the HOT
approach, but one I shall not press.
Suppose now that the HOT theory is correct. How could a scientist exploit it to infer a
conscious state from a subject's report? The subject reports that she is in M, a report
which is assumed to express a belief that she is in M. As long as state M exists, the
subject's belief or thought about it would be a higher‐order thought (M*), which entails
that M is conscious. Thus, the scientist can justifiably infer that state M is conscious. Of
course, the mere fact that the subject believes she is in state M does not, according to
the HOT theory, guarantee that she is. The HOT theory does not assume the infallibility
(invariable accuracy) of higher‐order beliefs. So there could be a belief that some state M
exists (at that time) although no such state does then exist (in the subject). As long as M
does exist, though, the HOT theory offers the subject's belief about M as conclusive
evidence for M's being conscious. In this fashion, the HOT approach looks promising for
the purpose of conferring warrant on a third‐person consciousness judgment.
I shall offer two criticisms of the HOT approach. First, it is unsatisfactory as a conceptual
analysis of state‐consciousness, as I shall show through a single type of problem. Second,
it does not cover all the territory that needs to be covered in terms of scientific
methodology. Half of what consciousness researchers do with verbal reports cannot beaccounted for via the HOT approach.
My critique of the HOT theory as a conceptual analysis focuses on its inability to explain
people's distinctive epistemic position vis‐à‐vis their own consciousness. When you are in
a conscious state M—at least a salient, nonfringy conscious state—then you automatically
have good evidence for believing that this state is conscious. 4 If you are consciously
thinking about Vienna, for example, that very thinking gives you excellent evidence for the
proposition that this thought is conscious. You don't have to believe the proposition, or
even reflect on it, to possess this evidence. 5 (If talk of “evidence” strikes the reader as
inappropriate, other phrases can be substituted. For example, (p.119) you have “good
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 6/22
Can Science Know When You're Conscious?
Page 6 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
reason” to believe the proposition, or you are in a “good epistemic position” to tell that it
is true.) This is a traditional philosophical view, and one that I believe is correct. Can it be
accommodated by the HOT approach? No, I shall argue.
Recall that the HOT theory is committed to the view that some conscious states are
rendered conscious by non‐conscious HOTs. Let us focus on that class of cases. Howdoes the HOT theory make sense of the fact that one automatically has good evidence for
the fact that these states are conscious? Since what makes them conscious is the
existence of independent, non‐conscious states, one could only have automatic evidence
for the consciousness of the first‐order states if one automatically had evidence for the
existence of those non‐conscious HOTs. 6 But why should such evidence automatically be
available?
A natural view is that a person has non‐inferential access to the consciousness of their
own conscious states. If this is right, then the HOT theory requires that one have non‐
inferential access to the existence of a non‐conscious HOT—because, in the class of casesunder discussion, the existence of such a non‐conscious HOT is what makes the first‐
order state have the property of being conscious. But non‐conscious states are precisely
ones whose existence can only be detected by inferential routes.
The situation is not much improved if we allow that access to one's own consciousness
may be inferential. It is still unclear where appropriate inferential evidence might come
from, especially in the present class of cases where the HOTs are non‐conscious.
Whether the person is thinking about Vienna or having a perceptual experience, what
enables her to infer that she simultaneously has a non‐conscious HOT directed at this
state? No inferential evidence seems to be standardly available to the subject about theHOT's existence.
Rosenthal might respond here that it is just a brute fact that one does have easy access,
whether inferential or non‐inferential, to the existence of non‐conscious HOTs. This is not
very satisfying. As Ned Block points out (personal communication), unconscious Freudian
beliefs are not easily accessible; for if they were, why would people pay psychoanalysts
large sums of money to uncover them? Similarly, unconscious perceptions are not easily
accessible, as is evident in many experiments. So it would be a most mysterious state of
affairs if the only non‐conscious mental states that are easily accessible are the HOTs that
make other states conscious.
I turn now to my second criticism of the HOT theory for the purposes at hand, viz., its
inadequacy in the role of vindicating scientists’ responses to verbal reports. At best the
HOT theory does only half the job that needs to be done. Though the theory can explain
why a scientist who receives a verbal report of M may legitimately draw the conditional
conclusion: “ If the subject is in M, then M is conscious,” it offers no explanation of why
the scientist (p.120) is entitled to draw the unconditional conclusion: “The subject is in
M, and M is conscious.” The report is assumed to express a HOT, and armed with this
HOT the scientist can ‘crown’ a suitable first‐order state with consciousness. But the
scientist must first ascertain that there exists a suitable first‐order state, and it has not
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 7/22
Can Science Know When You're Conscious?
Page 7 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
been demonstrated how the HOT theory enables him to do this. Assuredly, scientists
regularly do make inferences from “she reports she is in M” to “she is in M” (where ‘M’
is presumed to be conscious); but nothing in the HOT theory offers them a licence to
draw such inferences. So, in addition to its intrinsic inadequacy, revealed by its inability to
get the first‐person epistemology right, the HOT theory does not deliver the
epistemological goods that cognitive researchers need.
3. Functionalism and Verbal ReportsThe second version of the constitutive approach I shall consider is analytical functionalism.
According to analytical functionalism, adapted for present purposes, what it means for a
mental state to be conscious is that the state tends to give rise to a report of its content,
or to a belief that the state is present. In other words, as a first approximation, the
reportability or accessibility of a mental state is part of what it means for such a state to
be conscious. 7 If this approach is right, when reports of such states are observed this
constitutes (some) evidence that there is a conscious state of the type reported—
because that (supposedly) is what it is for there to be such a conscious state.
Alternatively, if the functional definition is phrased in terms of belief rather than report,
then as long as a scientist can infer that an appropriate belief has occurred this provides
evidence that there is a conscious state of the type reported—again, because that's what
it supposedly is for there to be such a conscious state.
Now according to almost every version of functionalism about consciousness, the
functional characterization is more complex than the foregoing requirement. Let us call
the foregoing requirement, formulated for belief, the “self‐intimatingness of conscious
states.” Self‐intimatingness implies that, for any conscious state M, being in M analytically
involves a disposition to believe that one is in M (at least for creatures with the
conceptual capacity to classify their mental states). No functionalist holds that the self‐
intimation condition exhausts the meaning of a state's being conscious. At most it would
be one conjunct in a larger conjunction of functional characteristics. Since reportability or
self‐intimatingness is only one conjunct among many, a scientist who observes a report
and infers a belief cannot logically deduce from them that the subject has a conscious
state of the sort reported. Still, the report would provide helpful evidence which could be
combined with other relevant evidence.
(p.121) I shall present two criticisms of this functionalist approach. First, I shall
challenge the adequacy of functionalism as a construal of how scientists use verbal
reports. Second, and more significantly, I shall suggest that even if a satisfactory
functionalist theory can be identified (a very questionable prospect), discovery of it will
rest on an independent, “first‐personal,” and nonfunctionalist understanding of
consciousness (roughly, “phenomenal” consciousness).
The first criticism parallels my second objection to the HOT approach. Functionalism
cannot adequately account for actual research practice in the handling of verbal reports.
As noted earlier, functionalism does not authorize a deductive inference from a subject's
belief or report that she is in conscious state M to the conclusion that she is in conscious
state M. The self‐intimatingness of conscious states only guarantees that one can make a
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 8/22
Can Science Know When You're Conscious?
Page 8 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
forward inference: from a person's being in conscious state M to her (probably) believing
that she is in M. It does not guarantee a backward inference from her belief (or report)
that she is in M to her actually being in M. But the backward inference is the central sort
of inference that scientists make and that we are interested in.
Let me put the matter in another way. According to a functional analysis of consciousnessthat includes a self‐intimation component, a person in conscious state M tends to believe
that she is in M. 8 However, this same belief (and report) might equally arise from many
alternative states that do not satisfy the full functional analysis of being in conscious state
M. Thus, a scientist who knows (on the basis of the subject's report) that she believes
she is in M is not warranted by functionalist considerations in inferring that she is in M
rather than one of the alternative states that might have given rise to the same belief (and
report). Yet it seems that consciousness scientists normally do take such reports as
ample warrant for the subject's really being in the reported state. 9
I turn now to my second criticism. Functionalism is fundamentally a third‐personapproach to mental states. Functionalism tries to characterize mental states and
properties, in this case consciousness, in purely causal terms, i.e., in terms of causal roles
that relate psychological states to external stimuli, physical behaviour and other
psychological states that are themselves characterized in terms of causal roles. None of
this is inherently first‐personal.
My own view, by contrast, is that consciousness is a phenomenon we initially understand
(in large measure) from a first‐person point of view. Any attempt to functionalize the
concept of consciousness, then, must try to get as close as possible to this phenomenon.
A satisfactory functional concept of consciousness must match the first‐person concepteither intentionally or at least extensionally. If some functional concept is presented that is
alleged to be a species of consciousness but turns out to be intentionally or (especially)
extensionally at odds with our first‐person concept of consciousness, the (p.122)
proper reaction, I submit, is to dismiss the functional concept as at best a concept of
shmonsciousness—some sort of related phenomenon—not a concept of consciousness.
Not everybody agrees with this methodological proposal. For example, Block and others
contend that there are multiple concepts of consciousness. 10 According to Block ( 1995 ),
there are at least two bona fide concepts of consciousness, including (1) a purely
functional concept, which he calls access consciousness, and (2) a non‐functional concept,
phenomenal consciousness. These are intentionally non‐equivalent (that is, there are
conceptually possible cases in which they diverge) and perhaps extensionally non‐
equivalent as well (there may be actual cases in which they diverge).
One problem with this approach is that indefinitely many functional concepts might be
adduced as candidate consciousness concepts. 11 Should we admit them all as genuine
varieties of consciousness, even when they conflict—possibly dramatically—with our first‐
person conception of consciousness? Shouldn't we rather say, as I urge, that most (if not
all) functional concepts are specifications not of consciousness but of some related
phenomenon (or phenomena): shmonsciousness? When a functional concept clashes with
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 9/22
Can Science Know When You're Conscious?
Page 9 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
our first‐person concept of consciousness, we should not conclude that it is a new form of
consciousness, e.g. X‐consciousness. Rather we should conclude that it isn't a
consciousness concept at all. There are passages and arguments in Block ( 1995 ) that
express this very point of view, but on the whole Block's approach seems rather
different, since it aims to legitimize a functional concept of consciousness that diverges at
least intentionally, and possibly extensionally, with the first‐person concept of
consciousness (phenomenal consciousness).
Let me highlight some of these points by looking more closely at Block's account of access
consciousness. He presents this concept as follows:
A state is access‐conscious (A‐conscious) if, in virtue of one's having the state, a
representation of its content is (1). . . poised for use as a premise in reasoning, (2)
poised for rational control of action, and (3) poised for rational control of speech. . . .
These three conditions are together sufficient, but not all necessary. ( 1995 , 231)
There are two initial problems with this proposal. First, since Block does not identify any
necessary conditions for A‐consciousness, the functional specification is importantly
incomplete. Second, several of the terms in the definition, especially “poised,”
“reasoning,” and “rational,” are open to many interpretations. Thus a large range of
rather different functional concepts might fall under this definition. Do all of them express
legitimate concepts of consciousness?
Block sometimes deals with this problem in ways quite congenial to my own favored
approach. Consider his use of the term “rational” in conditions (p.123) (2) and (3). Why
is this term inserted into these conditions? Block tells us to rule out the kind of controlthat obtains in blindsight (Block 1997 , 382). Block is saying, in effect, that if we omit the
rationality qualifier we shall be left with a functional concept that isn't a good sense of A‐
consciousness. Why isn't it? Because omission of the qualifier would allow informational
states distinctive of blindsight patients (i.e., states that produce only guessing behaviour)
to qualify as A‐conscious. Why would that be wrong? Presumably because it clashes with
our first‐person grasp of consciousness. When we imagine ourselves in the position of a
blindsight patient, we judge from that “pretend” perspective that there is no conscious
awareness of the visual information in question. This threatened mismatch with our first‐
person conception of consciousness induces Block to avoid a functional definition that
omits the rationality proviso. So Block acknowledges the need to bring the functional, or
access, concept of consciousness into alignment with the first‐person, or phenomenal,
concept of consciousness. I applaud this move as eminently appropriate; I merely
suggest that the same move be made across the board. When a functional concept
clashes with the first‐person concept of consciousness, the functional concept should be
abandoned as a concept of consciousness. (Of course, such a functional concept might still
have utility in psychological theorizing.)
There are other passages in which Block welcomes this policy. For example, he suggests
that the concept of A‐consciousness is “in some way parasitic” on the “core” notion of P‐
consciousness (Block 1995 , 274); and he offers the metaphor of a parquet floor that
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 10/22
Can Science Know When You're Conscious?
Page 10 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
requires another floor beneath it (274). Finally, he writes: “I should welcome attempts to
tinker with the definitions of ‘P’ and ‘A’ so as to make them coincide better” (277). When
he writes these things, he wisely endorses the sort of approach advocated here. But
when he urges upon us a functional notion of consciousness that avowedly departs from
the first‐personal notion, he veers off in a different, and to my mind misguided, direction.
Let us consider two further illustrations of this point. According to the research of Milner
and Goodale ( 1995 ), there are two types of visual processing systems in the human
brain: a dorsal pathway and a ventral pathway. The ventral stream processes visual
stimuli for the purpose of recognizing and identifying objects and events and for
constructing long‐term memories and models of the environment. The events in this
processing stream include conscious visual experiences. The dorsal stream, by contrast,
is specialized for processing visual input so as to control immediate action; it guides the
behaviour of reaching for objects and grasping them. The dorsal system, however, does
not involve consciousness. The existence of these two different systems yields, as an
upshot, two types of abnormal syndromes. When the ventral system is deprived of visual
input after damage to the visual area VI, patients have no conscious visual perception of
objects and events in the (p.124) world. Nonetheless, as long as the dorsal system is
spared, these patients can still accurately reach for and grasp objects by visual guidance.
Such a patient may grasp an object through visually obtained cues but will report no
conscious visual awareness of the object. In the converse syndrome (“optic ataxia”),
patients with a preserved ventral pathway but damage to the dorsal pathway can see
perfectly well in terms of visual awareness, but they are unable to effectively control
reaching and grasping behaviour.
In discussing these cases, Block remarks: “Though there is much disagreement about
the specializations of the two systems, it does appear that much of the information in the
ventral system is much more closely connected to P‐consciousness than that in the dorsal
system . . . . So it may actually be possible to damage A‐consciousness without P‐
consciousness and vice versa” (Block 1995 , 233). Although he doesn't give details, he
apparently thinks that the ventral system is a neural basis of P‐consciousness, the dorsal
system is a (partial) neural basis of A‐consciousness, and there could be one type of
consciousness without the other under conditions of damage of the sorts described.
Block's diagnosis strikes me as misguided, however. Why say that the dorsal system
involves any type of consciousness at all? Block apparently is thinking that damage to the
dorsal system constitutes damage to A‐consciousness because A‐consciousness, by
definition, partly involves the control of action. But here he clings to his functional
specification of A‐consciousness when it ought to be abandoned. Damage to the dorsal
system involves no impairment of consciousness at all, as Milner and Goodale's own
discussion of the case (tacitly) suggests. Milner and Goodale describe patients with optic
ataxia as retaining “conscious sight of the objects they are unable to reach towards”
(Milner and Goodale 1995 , 77). They rightly give no hint that damage to the dorsal
system creates any sort of consciousness impairment.
Another case worthy of attention concerns Block's term “poised,” which occurs in all
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 11/22
Can Science Know When You're Conscious?
Page 11 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
three conditions of his definition of A‐consciousness. “Poised” is so vague that a wide
variety of determinate functional specifications might fall under Block's definition. Which of
these does he intend? For example, are unretrieved but retrievable memory states
“poised” for use as premises in reasoning? Block explains that by “poised” he means
“ready and waiting” ( 1995 , 245), and he evidently intends this to exclude memory
states because they are not ready and waiting for use as premises in reasoning. I confess
that it isn't clear to me that unretrieved memory states fail to satisfy the phrase “ready
and waiting.” What is clear, however, is that any remotely acceptable functional
specification must exclude memory states. They must be excluded because, intuitively ,
unretrieved memory states are non‐conscious states. Moreover, it is the first‐personal
point of view that makes this classification transparent, not an antecedent grasp on some
functional characterization (p.125) of consciousness. Unretrieved memories simply do
not “feel” conscious; there is nothing “it is like” to have unretrieved information stored in
memory. Thus any functional specification that classifies them differently would be a bust
as a consciousness concept; at best it would be a shmonsciousness concept. Once againwe see how the first‐person perspective must guide and constrain any attempt at a
functional specification. At root, our epistemic window on consciousness is a first‐person
window. So, even if a functional specification of consciousness can ultimately be
constructed that is usable by consciousness researchers to infer subjects' awareness
states from their verbal reports, the vindication of this functional specification must
ultimately rest on a first‐person epistemological foundation (Block's underlying floor
beneath the parquet floor). 12
4. The Reliable Indicator ApproachThe second general type of approach mentioned in section 1 is the non‐constitutive
approach. This approach assumes that neither a person's report nor her underlying belief
that she is in conscious state M is (even partly) constitutive of her being in M, or of M's
being conscious. In fact, unlike the HOT theory and analytical functionalism, this approach
offers no particular analysis of consciousness. It just maintains that a belief that one is in
M is a reliable 13 indicator of actually being in M; at least this is assumed by
consciousness researchers. Subjects are regarded as epistemically competent to form
beliefs about the presence or absence of their conscious states, “competent” in the
sense that these beliefs are usually accurate.
There are various ways to qualify the claim of competence. If subjects form beliefs aboutthe micro‐structure of their conscious states, as they were invited to do in the
laboratories of historical introspectionists, such beliefs would not now be regarded as
reliable. Current cognitive scientists exercise caution in the sorts of questions they ask
their subjects to report about, using only familiar, folk‐psychological categories.
Nonetheless, experimenters often rely on subjects to follow instructions and make
responses about topics with considerable phenomenal complexity. A good example is the
work on mental imagery by Kosslyn and colleagues. When it comes to beliefs about the
absence of conscious states, researchers are particularly cautious. When subjects report
no awareness of stimuli close to a sensory threshold, or exposed for only a brief
duration, it may be argued that they were nonetheless aware of the stimulus.14
If we set
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 12/22
Can Science Know When You're Conscious?
Page 12 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
aside such cases, however, there is a strong tendency to regard reports of, and beliefs
about, conscious states as accurate.
One popular model of why (or how) people are competent about their conscious states is
the internal monitoring, or inner sense, model. 15 On this view, a person can form
accurate beliefs about their conscious states for the (p.126) same sort of reason thatthey can form accurate beliefs via the outer senses: they can somehow “scan” or
“monitor” these states. Philosophers commonly call such monitoring “introspection.”
There are two common objections to the inner sense model of self‐knowledge:
(1) Unlike cases of the outer senses, there does not seem to be any physical
organ associated with introspection.
(2) Unlike cases of the outer senses, there does not seem to be any distinctive
sensuous character associated with introspection.
The best reply to these objections, I think, is to point out that the inner sense theorysuggests only an analogy , an admittedly imperfect analogy, between introspection and
outer sense. So the existence of disanalogies can certainly be tolerated. Moreover, there
might well be an associated physical organ or mechanism, especially if you think that
introspection essentially amounts to attention. The physical mechanism would be
whatever the mechanism of attention is. Some proponents of the inner sense model use it
to define consciousness. I do not adduce the model in that vein, but merely as an attempt
to explain how introspective beliefs could be reliable indicators of the presence or
absence of conscious states. The beliefs are reliable indicators because they are
produced by a reliable mechanism or process.
Now, some people might argue that introspection is an infallible procedure—that it
necessarily yields accurate beliefs. Few contemporary philosophers endorse this once
popular view, and I agree with the contemporary consensus. First, it seems as if we
sometimes err in characterizing our mental states. In the dentist's chair we might think
that we are feeling pain when we are actually only feeling pressure. Furthermore, as long
as there is some temporal gap, however small, between a target mental condition and an
introspectively formed judgment about it, there seems to be room for forgetting, which
can generate error. Moreover, as long as there is even a tiny temporal gap, the
introspective judgment and the target condition must be “distinct existences.” If they are
distinct existences, how can there be a conceptually or metaphysically necessary
connection between them? Finally, as long as an internal monitoring mechanism is
conceived of as a physical device, it is presumably open to malfunction and therefore to
error (Lycan 1996 , 17). 16 Confabulation might be one such malfunction: either
confabulating the presence of a conscious state or confabulating its absence.
If introspection were infallible, i.e., necessarily reliable, the epistemological plight of
consciousness scientists would be considerably alleviated. But since introspection is at
best contingently reliable, as argued above, we face the question of what warrants
consciousness researchers in placing considerable trust in introspection's reliability. What
reason do they have for thinking (p.127) that introspection has 80% or 90% reliability,
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 13/22
Can Science Know When You're Conscious?
Page 13 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
or even reliability greater than chance?
Here is another way of posing the problem. If the present reconstruction is correct (as I
think it is), consciousness scientists tacitly rely on their subjects' introspective
mechanisms as instruments of their science. They rely on these mechanisms in the same
way scientists generally rely on instruments and measuring devices. But don't scientistshave to validate or calibrate their instruments before they use them? Don't they need
independent evidence to establish the accuracy of those instruments? What have
consciousness researchers done to validate the reliability of their subjects'
introspectors? Because they cannot directly observe their subjects' conscious states,
they cannot straightforwardly check up on the accuracy of the subjects' introspective
beliefs about these states. How, then, can these beliefs—or the verbal reports in which
they issue—legitimately be trusted?
The necessity of validating one's instruments is historically illustrated by Galileo's need to
convince his contemporaries of the telescope's reliability. Galileo was convinced that hehad discovered four “Medicean planets” (the moons of Jupiter) by means of the
telescope. But his opponents, steeped in Aristotelian philosophy, challenged the notion
that an instrument constructed from terrestrial materials would be accurate when used
to look at the heavens. So Galileo spent several years of hard work trying to demonstrate
to these opponents the telescope's reliability (Kitcher 1993 , 228–33).
Merikle and his colleagues have worried about the problem of validating the use of
subjective reports of awareness. Merikle's primary interest is in the issue of unconscious
perception, a topic that has elicited great debates over the proper procedures for
distinguishing conscious from unconscious processes. Investigators worry about theproper methodology to use in identifying thresholds of consciousness: at what thresholds
do perceived stimuli become conscious? One report criterion used here is to ask a
subject how confident she is that she can see a stimulus. The evidence for unawareness
is in terms of the subject's report of lack of confidence in her verbal judgment. The
trouble with this criterion, as pointed out decades ago by Eriksen ( 1960 ), is that it
places upon the individual subject the responsibility for establishing the criterion of
awareness. “We have no way of objectively knowing when [subject] A says a given
judgment was a pure guess and [subject] B makes a judgment and states it was a pure
guess that the two [subjects] are talking about the same phenomena or that they are
using the same criteria for guessing” (292).
Nonetheless, Cheesman and Merikle ( 1986 ) argue that a threshold defined in terms of
claimed or reported awareness is the best way to capture the distinction between
conscious and unconscious experiences; better, say, than a threshold of objective
chance‐level performance of discriminative responding. However, Cheesman and Merikle
do not want to leave the subjective (p.128) threshold unsupported or unvalidated,
because critics such as Holender ( 1986 ) claim that a purely subjective measure does
not rule out the possibility that partial stimulus information was consciously perceived. So
Cheesman and Merikle seek to find some converging, or corroborating, operation which
validates the subjective report criterion. They pursue this goal by trying to show that the
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 14/22
Can Science Know When You're Conscious?
Page 14 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
measure of subjective judgments of awareness is correlated with other, non‐subjective
measures, measures concerning behavioural effects associated with processing
perceptual information. This seems to be a search for the sort of independent validation
that we were discussing above. Merikle and colleagues successfully identify several
objective measures that involve qualitative differences between conscious and
unconscious influences as predicted by the subjective measure.
Here are two of these objective measures. It was found that (1) subjects have the ability
to initiate a predictive strategy involving a Stroop colour‐word priming task when and only
when they are consciously aware of the primes (Cheesman and Merikle 1986 ), and (2)
subjects' ability to successfully follow an exclusion instruction reflects greater conscious
than unconscious influences (Merikle and Joordens 1997a ). The exclusion task,
developed by Jacoby (Debner and Jacoby 1994 ), proceeds as follows. A five‐letter word
(e.g., “spice”) is presented and masked on each trial. Immediately following, a three‐letter
word stem (e.g., “spi—”) is presented and subjects are instructed to complete it with any
word except the one just presented. Debner and Jacoby ( 1994 ) found that subjects
were able to follow the instructions to exclude the immediately preceding words from
their completions when these words were presented for 150 milliseconds. However,
when the words were presented for 50 milliseconds, the subjects had difficulty following
the instructions and used many of the immediately preceding words to complete the
word stems, despite instructions not to use them. What is the supposed virtue of finding
these qualitative differences? As Merikle and Joordens express it:
Our search for converging evidence followed from previous suggestions that the
best way to validate a potential measure of awareness is to demonstrate that the
selected measure predicts a qualitative difference in performance across conscious
and unconscious perceptual states . . . . Not only can qualitative differences validate
a task as a measure of unconscious perception by constraining alternative
interpretations, but qualitative differences also serve to move the study of
unconscious influences forward by showing how conscious and unconscious
processes differ. In fact, it can even be argued that the distinction between
conscious and unconscious processes is of questionable value if conscious and
unconscious processes do not lead to qualitatively different consequences. ( 1997b
, 115–16)
My problem with these findings of qualitative differences concerns the theoretical
conclusions to be drawn. We started out worrying whether subjects' (p.129)
introspective reports about their conscious states are accurate. Merikle's various
findings show that subjects' reports are not isolated or random events; rather, they
correlate with performance on other tasks. To my mind, this tends to show that the
subjects' verbal reports are veridical signs of something, of some bona fide psychological
events or processes. But why do they show the presence or absence of consciousness
(or awareness)? Unless the objective tasks—initiating a predicting strategy or following
an exclusion instruction—are known independently to be good measures of
consciousness, the fact that they correlate with subjective reports goes no distance
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 15/22
Can Science Know When You're Conscious?
Page 15 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
towards showing that the latter are accurate measures of consciousness. But how is it
determined independently that the objective tasks are good measures of consciousness
vs. unconsciousness?
This problem is tacitly acknowledged by Cheesman and Merikle. They defend the
assumption that the ability to initiate a predictive strategy is a good measure of consciousness by saying: “This assumption is consistent with views such as those
expressed by both Posner and Snyder . . . and Underwood . . . who equate the attentional
processes underlying strategy effects with consciousness” (Cheesman and Merikle 1986
, 347–48). This just raises the same problem once again, however! How can researchers
have established the relationship between consciousness and attentional processes
(defined in some empirical manner) without a prior method of ensuring the presence of
consciousness? I f the prior method relies on verbal reports, we are moving in a circle.
Similarly, how can researchers have demonstrated a relationship between consciousness
and an ability to follow an exclusion instruction without a prior method of determining the
presence of consciousness? I f this prior method relies on verbal reports, we are again
caught in a circle.
I would formulate the general theoretical problem as follows. In the case of arbitrary
constructs in psychology, one might be interested in whether a given criterion measures
anything objective of interest. One seeks a converging measure systematically associated
with the target measure in order to corroborate the fact that there is some objective
phenomenon that both are measuring. But in the present case we have a pre‐theoretical
phenomenon, consciousness, and we are wondering whether a proposed measure—viz.,
subjective beliefs or reports—is a good measure of that phenomenon. Furthermore,
under the non‐constitutive approach, we are assuming that it is not a conceptual or
definitional truth that verbal reports are reliable indicators of consciousness. Nor is it a
definitional truth, presumably, that any objective task performance is a reliable indicator
of consciousness. So how are we supposed to identify any empirically observable variable
that validates verbal reports? Whatever scientific variable, V, is found to correlate with
verbal reports, this will not help validate the reports as a measure of consciousness
unless V itself is an accurate measure of consciousness. But how can that be ascertained
without relying on verbal reports?
(p.130) 5. Starting from the First Person A first step towards a possible resolution of the problem is to give a fuller reconstruction
of what scientists tacitly assume in relying on verbal reports. I conjecture that
consciousness researchers, when relying on their subjects' reports, tacitly appeal to
their own knowledge of their conscious states. In effect, they appeal to a sort of analogical
argument: “I have reliable knowledge of my own states of consciousness; therefore it is
reasonable to assume that other people have reliable knowledge of their states of
consciousness.” Deployment of this sort of analogical argument would be of a piece with a
certain theory in the “mind‐reading” literature, viz., the simulation theory (Goldman
1989c ). The simulation theory says that a principal method one uses in attributing mental
states to others is to “model” others' mental states by trying to instantiate them, in
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 16/22
Can Science Know When You're Conscious?
Page 16 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
pretend mode, in oneself. Tacitly, this relies on an assumption that others are like oneself.
Although I do not suggest that consciousness scientists always deliberately proceed in
this simulationist mode, their study of consciousness may well be founded on some such
assumption of similarity‐to‐self.
Some evidence for this procedural assumption is the fact that cognitive psychologists liketo try out experiments on themselves. Occasionally their reliance on personal experience
is even made explicit, as in a recent book by Baars ( 1997 ). Baars imagines a sceptic who
wonders whether science has anything to do with real consciousness. He writes:
The best reply I can think of is to ask sceptics to try one of the demonstrations in
this chapter, and ask, is this truly your experience? If yes, an honest sceptic should
say that we are indeed dealing with genuine consciousness as a valid topic for
exploration. . . .
Could the evidence regarding consciousness just be a clever imitation of the realthing? . . . As an answer, we need only notice that consciousness as an object of
scientific scrutiny fits our personal experience remarkably well. That is not likely to
be a coincidence. (34)
In this passage Baars suggests that the foundation of the scientific study of consciousness
rests on first‐person “observation.”
A serious question still remains, however. Is reliance on first‐person observation
scientifically legitimate? If first‐person observation essentially consists in introspection,
aren't we still stuck with the problem of validating introspection's reliability? That was theproblem we identified for researchers' reliance on their subjects' introspection. How is
the problem in any way alleviated by showing that they rely on their own introspections?
Isn't the reliability of introspection still in need of independent validation?
Some epistemologists would certainly insist on such validation. For example, BonJour (
1985 ) contends that a necessary condition for a belief to (p.131) be justified is that the
believer have a metajustification for that belief. What that amounts to, for present
purposes, is that an introspectively formed belief cannot qualify as justified unless the
believer has a justification—presumably an independent justification—for believing that
introspection is reliable.17
If an independent metajustification is indeed insisted upon,that is a difficult condition to meet. But is it clear that a proper epistemological theory
should impose such a condition? No. Many epistemologists, myself included, think that
this independent validability requirement is too restrictive or demanding. In particular,
when it comes to one's own basic cognitive processes, it would be too much to require
that each process be validatable in an independent fashion. Taking such a path leads
straight to global scepticism (as I sketch below). This does not mean that independent
validation should never be required. It does not mean that it isn't an appropriate
requirement for scientific instruments external to our natural cognitive equipment. All it
means is that an independent validation requirement should not be imposed “all the way
down.” 18
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 17/22
Can Science Know When You're Conscious?
Page 17 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
To illustrate the scepticism‐inducing potential of such a systematic requirement, consider
the case of memory. As Alston ( 1993b , 116–17) argues convincingly, it is impossible to
establish (validate) the reliability of memory without appealing to premises that
themselves rest on memory. Non‐circular validation might seem possible, because one
can check up on whether an ostensibly remembered event really occurred by consulting
records or traces of past events. But how can those records be trusted without in turn
relying on memory? How can I trust what an old New York Times says about the past
without relying on memory to support the trustworthiness of the New York Times (and
without relying on memory to support the belief that the ink‐marks on its pages express
such‐and‐such contents)? Though the circle may be large or small, there is no non‐
circular way of validating memory. Ultimately, memory is our only direct avenue to the
past, our only cognitive resource that provides an epistemic window on the past. If we
allow memory beliefs to qualify as justified only on the condition that memory's reliability
be established in an independent, non‐circular fashion, then we will have no justified
memory beliefs. The same sort of argument applies to perception. As I argued in chapter5 , an unqualified independent validation requirement would imply that we have no
justified perceptual beliefs (see also Alston 1993b ). For these reasons, a systematic,
unqualified form of such a requirement must be rejected. When we reject this
requirement we also free introspection from its yoke. Introspection can yield justified
beliefs even if a user cannot perform the impossible feat of giving an independent
demonstration of introspection's reliability.
I do not contend that there are no further constraints that should be imposed on the
legitimacy or viability of trusting introspection. For example, although it may be
impossible to validate or invalidate introspection by entirely independent (p.132)means, one might demand that introspection pass the test of self‐consistency. That is, it
should not deliver inconsistent judgments about conscious states, or at least it should not
deliver too many inconsistent judgments, on pain of proving itself to be highly unreliable.
To be sure, passing the test for rampant inconsistency would not establish reliability; but
passing such a test could at least be considered an epistemological “plus.” If memory
frequently delivered inconsistent beliefs, we would be forced to infer that it is not very
reliable (even if we couldn't tell which memory beliefs were false). It is presumably a
favourable epistemic feature of memory that a very large number of memory beliefs
cohere with one another. The same would be true for introspection.
We can also demand that introspection should pass the test of not yielding too many false
beliefs (as far as we can tell) when combined with other presumptively reliable processes
or procedures. Here is an example of how introspection can pass such a test. I now have
an introspectively formed belief that I currently intend to snap my fingers in the next two
minutes. I have a background belief that intentions usually stick around for a while and
get acted upon if they are not overridden. I therefore predict that I shall indeed snap my
fingers in the next two minutes. If that action is then observed within the specified time
interval, this partly “corroborates” the reliability of introspection. That is, it corroborates
introspection's reliability in the weak sense of not defeating or not challenging
introspection's reliability. There was a chance that introspection (together with other
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 18/22
Can Science Know When You're Conscious?
Page 18 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
processes) should have yielded a false belief, but, assuming I can trust my visual
observation, this did not occur. So introspection survived a test of sorts. This might be
re‐described by saying that introspection coheres—substantially, at any rate—with other
belief‐forming methods. Passing such a coherence test might be regarded as a
necessary, though not a sufficient, sign of its reliability. 19
Historically, of course, it was thought that introspection was shown to generate massive
contradictions, thereby revealing its substantial unreliability. In that period, however,
introspection was expected to reveal details of mental life that current cognitive scientists
would not demand of it. Maybe it was proved that introspection isn't reliable on the old,
overly ambitious range of questions; but this doesn't establish that it is unreliable when
applied to a narrower range of questions. I submit that introspection has not been proved
to be unreliable with respect to “macroscopic” conscious states, such as whether I now
consciously intend to snap my fingers, whether I am now consciously thinking about
Vienna, or whether a particular part of a stimulus looks to be figure or ground. To specify
more exactly the range or scope of introspection's reliability is a delicate matter. This is
not the place to try to embark on such a project. All I am presently arguing is that
introspection seems to pass the sorts of tests that one can reasonably demand that it
pass. This suggests that introspection does not fail any appropriate tests (p.133) for
epistemological admissibility, contrary to the sorts of worries examined earlier. 20
Thus far I have been addressing the justifiability of regarding one's own introspective
faculty as reliable. What about the third‐person case? Can one legitimately infer the
reliability of other people's introspectors from the reliability of one's own? The first point
to notice is that one can utilize the same sort of coherence test for other people's
introspectors as discussed above for one's own introspector. If someone tells me that
she now intends to snap her fingers in the next two minutes, I can do at least a coherence
test of this (ostensible) introspective report by seeing whether she does snap her fingers
within two minutes. If such introspective reports typically comport with other observable
behaviour, this tends to “corroborate” the reliability of her introspector. But why should
I regard other people's verbal reports as reports of consciousness at all? Isn't that the
question we posed originally? In discussing Merikle's work, moreover, we challenged the
notion that other tasks people perform (apart from giving verbal reports) can confirm
verbal reports as evidence for consciousness. How does the present appeal to other
behaviour, however “coherent” it might be with verbal reports, help resolve this issue?
At this juncture I think we need to return to the analogical inference from one's own case.
What makes it plausible that other people's verbal reports are expressions of (beliefs
about) conscious states (or their absence) is, in large measure, the apparent similarity of
other people to oneself. This is confirmed by the way we intuitively address questions
about the consciousness of animals. We tend to be most confident of the consciousness of
animals most similar to ourselves, and less confident as they appear more dissimilar. 21
Moreover, at least a scintilla of doubt remains in many people about even the most similar
animals precisely (it would appear) because of their dissimilarity from us.
Even if these remarks support the similarity‐to‐self idea as the basis for our intuitive
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 19/22
Can Science Know When You're Conscious?
Page 19 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
judgments of consciousness, do they help with the question of the scientific admissibility
of verbal‐report‐based research on consciousness? Is there any epistemological
legitimacy in such analogical inferences? Philosophers of mind, as a rule, do not take kindly
to the argument from analogy to other minds, but I find that other thinkers find this sort
of approach congenial. One published endorsement of it is by a noted economist and
game theorist, Harsanyi, who discusses it in connection with interpersonal utility
comparisons (Harsanyi 1982 ). Harsanyi says that interpersonal utility comparisons rest
on what he calls the “similarity postulate.” This is the assumption that, once proper
allowances are made for differences in taste, education, etc., between me and another
person, it is reasonable for me to assume that our basic psychological patterns are much
the same. Harsanyi calls this a nonempirical a priori postulate and contends that such
postulates are common in science. He specifically compares it to a priori, nonempirical
criteria of (p.134) theory choice such as simplicity, parsimony and preference for the
“least arbitrary” hypothesis. We might add to this list the preference in science for
theories and models displaying symmetries. Van Fraassen ( 1989 ) formulates thispreference in the form of the following slogan, which he calls the “symmetry
requirement”: “Problems which are essentially the same must receive essentially the
same solution” (236). He paraphrases this as the principle, “respect the symmetries”
(236). If this is an appropriate principle for physical science (from which all of van
Fraassen's examples are drawn), why should it not be appropriate for cognitive science?
This may be the best strategy for consciousness scientists to take in defending the
scientific mettle of their procedures. It may provide an epistemically respectable route
from a first‐person grasp of consciousness to the third‐person study of consciousness.
6. ConclusionI present the foregoing proposal in a tentative, groping fashion, not as a definitive or fully
defended solution to this nest of problems. The entire chapter, in fact, is mainly presented
as a reminder that the epistemological dimensions of consciousness research are just as
difficult and daunting as the metaphysical ones, on which most of the recent philosophical
discussion has focused. I have tried to show that the epistemological credentials of
consciousness research cannot be established by trying to put such research on a
purely third‐person footing; the first‐person grasp of consciousness remains fundamental.
Nonetheless, this need not abort the prospects for the methodological respectability of
consciousness research, as long as one does not succumb to the temptation of placing
excessive constraints on epistemic justification.
NotesThanks to Ned Block, David Chalmers, David Rosenthal and Holly Smith for very helpful
comments on earlier drafts of this chapter, and to the editor and two anonymous
referees for constructive suggestions.
Notes:
(1.) In this usage I follow many (though not all) practitioners. For example, Weiskrantz
titles a paper of his “Some Contributions of Neuropsychology of Vision and Memory to
the Problem of Consciousness” (Weiskrantz 1988 ), but employs the terms “aware” and
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 20/22
Can Science Know When You're Conscious?
Page 20 of 22
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2014.All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy ). Subscriber: OUP - OxfordOnli ne %28Sales %26 Publicity%29; date: 14 Apri l 2014
“awareness” in most of the text. In two papers on prosopagnosia, Tranel and Damasio (
1985 , 1988 ) use “awareness” in the title of one of them and “conscious” in the other.
(2.) Although Rosenthal says that he offers the HOT theory as a “hypothesis” rather than
a “definition” (Rosenthal 1999, 6), his various ways of formulating the matter all make it
sound rather close to a definition. He presents his account as saying “what it is in virtue
of which conscious mental states are conscious” (2, emphasis added); “for a mental state
to be conscious is for one to be conscious that one is in that very state” (2); “a mental
state's being conscious consists not just in one's being conscious of that state, but in
one's being conscious of it under some relevant description” (3, emphasis added).
Moreover, Rosenthal's own stated view notwithstanding, it seems most natural to
understand the HOT theory as a theory about the meaning of state consciousness.
(3.) Rosenthal explicitly accepted this point when queried about it at the third annual
meeting of the Association for the Scientific Study of Consciousness, London, Ontario, 5
June 1999.
(4.) Also, the fact that a state is conscious implies that it is an easy object of potential
cognition (by the subject). This must be distinguished, however, from the core thesis of
the HOT approach, viz., that a state's being conscious implies that there is some actual
(contemporaneous) cognition of it.
(5.) Being in this epistemic position does depend, however, on possession of the concept
of consciousness. Creatures lacking the concept of consciousness might be in certain
conscious states without having good reason to believe that they are, simply because
they lack the concept.
(6.) It might be argued that my objection does not cut against Rosenthal's own theory
because his is not a semantical‐conceptual account of consciousness. If “all Xs are Ys”
expresses a semantical truth, as in “all bachelors are unmarried,” then indeed one cannot
have evidence for something being an X without having evidence for its being a Y. But
when “all Xs are Ys” expresses merely a contingent truth, as in (perhaps) “all water is
H2O,” one can have evidence for something being an X without having evidence for its
being a Y. So my argument does not hold against Rosenthal's own version of the HOT
theory. However, if Rosenthal sticks to a non‐semantical, non‐conceptual version of the
HOT theory, he owes us a different theory of the meaning of (state) consciousness, andmore of an explanation as to why, given this different meaning of “consciousness,” it
nonetheless turns out that state‐consciousness consists in having a suitable HOT.
(7.) Versions of functionalism that express these themes can be found in Putnam ( 1960 ),
Dennett ( 1969 ), and Shoemaker ( 1988 ).
(8.) One might want to build more conditions into this account; for example, her being in
M will tend to produce a belief that she is in M only if she first attends to the question of
whether she is in M. I neglect this type of consideration in the interest of simplicity.
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 21/22
8/10/2019 Can Science Know When Youre Conscious
http://slidepdf.com/reader/full/can-science-know-when-youre-conscious 22/22
Can Science Know When You're Conscious?
Page 22 of 22
some changes and improvements in my general account of justifiedness in Goldman (
1992 , ch. 9).
(19.) Since I am only adducing coherence as (at most) a sign of reliability, I do not regard
the present proposal as a way of transforming myself from a reliabilist into a coherentist.
Indeed, in earlier writings I also adduced coherence as a useful guide to, or test for,reliability (see Goldman 1986 , 100).
(20.) Some might object that I have not confronted the central objection to introspection
as a procedure, viz., its privacy . It is commonly maintained that science, at least, should
only use public methods. I have addressed this point at length in chapter 5 . I criticize the
publicity assumption by showing, through examples, that methods don't have to be public
to qualify as scientific. Admittedly, the examples adduced are mostly hypothetical or
“thought‐experimental” in nature. But such examples are legitimate when we seek to
clarify what is or isn't required by our very conception of science.
(21.) Of course, much depends on the respects of similarity. I am unable to say which
respects of similarity loom largest.
Access brought to you by: OUP - Oxford Online %28Sales
%26 Publicity%29